ORLANDO, Fla. -- Server virtualization allows much higher rates of system usage, but the resulting increases in network traffic pose significant challenges for enterprise storage. The simple "single server, single network port" paradigm has largely been displaced by servers running multiple workloads and using numerous network ports for communication, resiliency and storage traffic.
Virtual workloads are also stressing storage for tasks, including desktop instances, backups, disaster recovery (DR), and test and development.
In 2012, 58% of installed x86 workloads are running in a VM. By 2015, 77% of installed x86 workloads are expected to be running in a VM
At Gartner Symposium/ITxpo here this week, Stanley Zaffos, a Gartner research vice president, outlined the implications of server virtualization on storage and explained how storage virtualization solutions, the right approach, and the proper tool set can help organizations mitigate the impact on enterprise storage.
Consider using storage virtualization. Zaffos urges organizations to deploy storage virtualization as a means of better storage practice, and he underscores core benefits of the technology:
- Storage virtualization supports storage consolidation/pooling, allowing all storage to be "seen" and treated as a single resource. This avoids orphaned storage, improves storage utilization and mitigates storage costs by reducing the need for new storage purchases. The benefits of storage consolidation increase with the amount of storage being managed.
- Storage virtualization supports agile and thin provisioning, allowing organizations to create larger logical storage areas than the actual disk space allocated. This also reduces storage costs because a business does not need to purchase all of the physical storage up front -- simply add more storage as the allocated space fills up. Later tools may allow dynamic provisioning where the logical volume size can be scaled up or down on demand. Management and capacity planning is important here.
- Storage virtualization supports quality of service (QoS) features that enhance storage functions. For example, auto-tiering can automatically move data from faster and more expensive storage to slower and less expensive storage (and back) based on access patterns. Another feature is prioritization, where some data is given I/O priority over other data.
Consider using solid-state drives (SSDs). One of the gating issues for storage is the lag time caused by mechanical delays that are unavoidable in conventional hard-disk technologies. This limits storage performance, and the effects are exacerbated for virtual infrastructures where I/O streams are randomly mixed together and funneled across the network to the storage array, creating lots of disk activity. Storage architects often opt to create large disk groups. By including many spindles in the same group, the mechanical delays are effectively spread out and minimized because one disk is writing/reading a portion of the data while other disks are seeking. Zaffos points to SSDs as a means of reducing spindle count and supplying much higher IOPS for storage tasks.
Plan the move to virtualization carefully. Data center architects must develop a vision of their infrastructure and operation as they embrace virtualization. Zaffos suggested IT professionals start by identifying and quantifying the impact server virtualization, data growth and the need for 24/7 operation will have on the storage infrastructure and services.
The pace of virtualization has not slowed. More VMs have been deployed in 2011 than in 2001 through 2009 combined
Next, determine what you actually need to accomplish and align storage services with the operational abilities and physical infrastructure. For example, if you need to emphasize backup/restoration capabilities, support data analytics, or handle desktop virtualization, it's important to be sure that the infrastructure can support those needs. If not, you may need to upgrade or make architectural changes to support those capabilities.
When making decisions for virtualization, Zaffos notes the difference between strategic and tactical issues. Strategic decisions create lock-in, and tactical decisions yield short-term benefits. For example, the move to thin provisioning is a tactical decision, but the choice to use replication like SRDF would be a strategic decision.
Use available tools. Storage virtualization benefits from tools that can optimize storage activity. Zaffos points to examples including VMware Storage I/O Control (SIOC) and vStorage APIs for Array Integration (VAAI). SIOC essentially provides storage administrators with I/O and bandwidth throttling, along with I/O prioritization. This allows administrators to allocate resources to high-priority workloads and minimize the effects of other I/O-heavy (but less critical) workloads. By comparison, VAAI allows storage developers to integrate array products with vSphere, enabling features like full copy, block zeroing and hardware-assisted locking.
Ultimately, Zaffos notes that storage virtualization solutions can be a key enabling technology for server and desktop virtualization -- both of which place extreme demands on the storage infrastructure. But, he said, the move to storage virtualization takes a thorough understanding of the benefits, careful planning to ensure proper alignment with business and technical needs, and judicious use of storage technologies like tiers and SSD.